J Econ Finan DOI 10.1007/s12197-013-9265-z

The linkage between the U.S. “fear index” and ADR premiums under non-frictionless stock markets Omar A. Esqueda & Yongli Luo & Dave O. Jackson

# Springer Science+Business Media New York 2013

Abstract This paper examines the effects of the U.S. investor sentiment on American depository receipts (ADR) premiums by using daily prices of Latin American ADRs from 1995 to 2009. The volatility index (VIX) is used as a proxy for investor expectations about the stock market. High levels in the VIX indicate that investors are fearful about future performance of the U.S. stock market. We estimate a GARCH-M in the framework of an ADR pricing model. We control for liquidity, transaction costs, and domestic and U.S. stock exchange returns. We find that deviations from the law of one price in ADRs can be partially explained by the lag of the smoothed volatility index. There is a structural break in the sample period before and after the enactment of the Sarbanes-Oxley Act. This paper has important implications for portfolio diversification on emerging economies as investment managers can improve hedging strategies by incorporating known values of the volatility index. Keywords American depository receipts . ADR premium . Latin America . VIX . Implied volatility . Investor sentiment . Law of one price JEL Classification G14 . G19 We are grateful to participants of the 47th AEF Annual conference for relevant suggestions on an earlier version of the manuscript. We also thank participants of the 50th SWFA annual meeting in Houston, TX for constructive comments. We recognize Dr. Diego Escobari, Dr. Emilios Galariotis, and Dr. David Johnk for valuable inputs. We are grateful to an anonymous referee for important suggestions. All remaining errors are our own. O. A. Esqueda Department of Accounting, Finance, & Economics, Tarleton State University, 1333 West Washington Street, Stephenville, TX 76402, USA Y. Luo School of Business, Wayland Baptist University, 1900 West 7th Street, Plainview, TX 79072, USA D. O. Jackson Department of Economics & Finance, The University of Texas-Pan American, 1201 West University Drive, Edinburg, TX 78539, USA O. A. Esqueda (*) T-0920, Stephenville, TX 76402, USA e-mail: [email protected]

J Econ Finan

1 Introduction American depository receipts (ADRs) have become one of the most important financial instruments for companies from emerging economies to overcome capital barriers by accessing U.S. stock markets. Additionally, ADRs have been commonly used as convenient instruments for international diversification as they are available in the U.S. and denominated in U.S. dollars. Consequently, the breadth and availability of ADRs has been constantly increasing as more firms continue cross-listing their shares in the U.S.1 According to the law of one price (LOP), the ADR price must equal the price of the underlying security after adjusting for the ADR ratio and exchange rate. However, there seems to be a systematic mispricing, or ADR premium, in the ADR market. We attempt to identify the sources of the ADR mispricing. One of the most important limitations to calculate the ADR premium is the lagged effects. For instance, there is a significant timing difference between Asian and North American stock exchanges. However, Kim et al. (2000) find that the effect of stock markets and exchange rate fluctuations on ADR prices is mostly completed during the same calendar day. Unlike their Asian and European counterparts, Latin American countries share a high and homogeneous linkage to the U.S. financial markets. In addition, simultaneous trading hours also have the beneficial effect of reducing the likelihood of lagged effects. Our findings on ADRs from Latin American countries can be extended to other emerging market ADRs. The outstanding growth of emerging economies has been reflected in the high number of ADR programs initiated in recent years. Figure 1 depicts the number of exchange-traded ADR programs by region and indicates that Latin American countries comprise a substantial portion of the ADR market. Using 69 ADR programs from Argentina, Brazil, Chile, and Mexico over the period from January 1995 to May 2009, we test whether investor sentiment has a significant impact on ADR premiums. We use the Volatility Index (VIX) as a proxy for investor sentiment. High levels in the VIX indicate that investors are likely to have pessimistic expectations about the U.S. stock market.2 We define the ADR premium as the disparity between the ADR price and the price of the underlying share, after adjusting for the corresponding exchange rate and ADR ratio.3 In addition to the aforementioned factors, we control for the domestic and the U.S. stock market indices which are widely documented as relevant determinants of ADR prices in the literature (Kim et al. 2000; Fang and Loo 2002; Bin et al. 2004; Bae et al. 2008; Alhaj-Yaseen 2011; Esqueda and Jackson 2012). We find a negative and significant relationship between investor sentiment and ADR premiums. The deviations from the LOP in ADRs can be partially

1

For instance, the number of type 2 and type 3 ADR programs from Latin America rose from 22 in 1994 to 81 by the end of 2010. Type 2 and type 3 ADR programs are traded on U.S. Exchanges and they are the most restrictive types of programs. Additionally, they tend to be the largest firms in their home country. Type 1, over-the-counter (OTC), and type 4, Rule 144-A ADRs, are excluded due to data constraints. 2 The VIX is quoted in percentage points and translates, roughly, to the expected movement in the S&P 500 index over the next 30-day period. The index is a weighted average and is quoted in real time by the Chicago Board Options Exchange. 3 ADR ratio refers to the number of underlying shares that one ADR can be converted to. When an ADR program is available, investors holding underlying shares can convert them into an ADR by paying the applicable fees. Investors might also choose to convert ADR into firms’ underlying stock.

J Econ Finan

Number of Cross-Listed Firms by Region 120

100

80

60

40

20

0 Western Europe

Latin America

Mainland China Asia Excl. China Eastern Europe

Other

Fig. 1 Number of cross-listed firms by region. ADR levels 2 and 3 as of December 2010. Source: Citigroup depositary receipt services

explained by investors’ fears about the future of the stock market, as evidenced by changes in the volatility index. Additionally, current U.S. stock market behavior is also a relevant pricing factor. We conclude that investors can improve their hedging strategy by incorporating the lagged values of the volatility index in determining ADR prices. This paper contributes to the existent literature in the following ways. First, we relate ADR premiums to the volatility index, which measures investors’ “fear” about future stock market movements. To test the statistical inference between ADR premiums and investor sentiment, we use a generalized autoregressive conditional heteroskedasticity in mean model (GARCH-M) and acknowledge the potential conditional variance of the ADR premium in our sample period due to country events such as currency depreciations. Second, we control for the relevant factors in ADR pricing by comparing the response of ADR premiums to changes in the volatility index. The results have important implications for constructing country portfolios by observing investors’ preferences.4 In the next section, we review the existing literature. Section 3 describes the data and sampling procedure. In section 4, we explain the econometric techniques employed and address our research questions. Section 5 presents a detailed description of the empirical results and Section 6 concludes.

2 Literature review Previous literature documents that ADR premiums are commonly found and their determinants are puzzling. According to the LOP, two assets with similar risks cannot

4

The underlying stock, the domestic and U.S. indices, and the foreign exchange rate are important pricing factors according to Aquino and Poshakwale (2006). Kim et al. (2000) only consider exchange rate, the domestic market conditions, and the U.S. market conditions.

J Econ Finan

be priced differently, hence, in equilibrium, the ADR premium must be equal to zero. This implies that investors should be able to invest in ADRs without being concerned about potential ADR mispricing arising from stock-market conditions, firm-specific events, or exchange-rate fluctuation. However, such price stability is often not supported by the existing literature. For instance, Jackson and Madura (2003) find arbitrage opportunities in the ADR market following profit-warning events. Suarez (2005) also shows that French ADRs are commonly mispriced and that arbitrage opportunities exist. Later, Grossman et al. (2007) find that deviations from the LOP in ADR can be the consequence of investors’ sentiment and/or transactions costs. The literature generally puts forward a very limited number of factors as determinants of ADR prices. Aquino and Poshakwale (2006) suggest that the innovations in the underlying stock, domestic and U.S. indices, and the changes in exchange rate are all important pricing factors; in addition, they find that exchange rates have a negative correlation with ADR returns. Kim et al. (2000), in an ADR pricing model, incorporate the relevant exchange rate, the domestic market conditions, and the U.S. market conditions. We concede that ADR prices should adjust for exchange rate by simply converting share prices to U.S. dollars, since Liang and Mougoué (1996) and Fang and Loo (2002) suggest that the exchange-rate risk in ADRs is diversifiable and does not command a premium. However, Bae et al. (2008) find that investors demand a premium due to exchange-rate risk and that such premium differs among U.S. and local investors. Other authors have put forward evidence suggesting that exchangerate risk results in considerable losses for ADR investors (Bin et al. 2004; Esqueda and Jackson 2012). We analyze whether ADR disequilibrium exists and whether it is motivated primarily by investors’ sentiment and transaction costs. Investor sentiment is a relatively new area in behavioral finance. Barberis and Thaler (2003) posit that behavioral finance is derived from two major aspects; psychology and limits to arbitrage. In finance, investors’ psychology is often used interchangeably with investors’ sentiment. In the area of the limits to arbitrage, Shleifer and Vishny (1997) conduct the first thorough study. The authors state that asset prices’ deviations from fundamentals is only a necessary condition for arbitrageurs to profit and bring the price back to fundamental levels. In addition, transaction costs play an important role, as potential arbitrage profits need to more than offset these costs. If transaction costs are at least as high as arbitrage profits, then market efficiency holds (Fama 1970). However, Shleifer and Vishny (1997) suggest a limited role for arbitrageurs due to a mythical riskless arbitrage. Prices might be away from fundamentals for a longer period of time than the arbitrageur can hold the assets, assuming that the arbitrageur does not have infinite resources. Classic textbook cases of riskless arbitrage profits are almost non-existent. In non-frictionless markets, arbitrage opportunities are inherent with some type of risk. Hence, costly arbitrage limits arbitrageurs from intervening in the market when prices deviate marginally from fundamentals. When attempting to earn arbitrage profits, investors must be concerned with transaction costs and holding costs (Pontiff 1996). If the risks are permanent or long-lasting, price deviations away from fundamentals are possible. Nonetheless, a small level of irrationality in the market (“noise” trading) can drive the price away from fundamentals, under strategic complementarity, but not under strategic substitutability (Fehr and Tyran 2005)

J Econ Finan

Several proxies for investor sentiment have been applied in previous research. Lee et al. (1991) suggest that the closed-end fund discount is a good measure of investors’ market expectations, where high discounts suggest bear markets. Baker and Wurgler (2006, 2007) use six proxies for investor sentiment: closed-end fund discount, dividend premium, initial public offerings (IPOs) volume, first-day return on IPOs, New York Stock Exchange (NYSE) volume, and equity share in new issues. Based on these proxies, they develop a composite measure of sentiment which is adjusted for production cycles. However, this measure of sentiment cannot be attributed to a particular type of investor and is computed on a monthly frequency. They conclude that stocks difficult to price appear more vulnerable to investor sentiment, probably because arbitrageurs stay away from these types of stocks. DeLong et al. (1990) suggest that “noise” traders can drive prices away from fundamentals, however, informed traders are able to profit if they purchase (sell) ahead of the “noise”-related demand (supply). In this scenario, informed-traders do not drive prices back to equilibrium levels, as would be expected in the finance literature. Until recently, most researchers treat sentiment measures as irrational or “noise” trading (Brown and Cliff 2004). Fisher and Statman (2000) use the Merrill Lynch Global Fund Managers Survey as a proxy for institutional investor sentiment, and the Investor Intelligence (II) survey for semi-professional investors, such as finance magazine writers. Brown and Cliff (2004), Verma and Soydemir (2006, 2009), and Verma and Verma (2008) measure sentiments using the II survey for institutional investors’ sentiment and the American Association of Individual Investors (AAII) for individual investors’ sentiment. Finally, Dennis and Mayhew (2002) use the put-call ratio to gauge investors’ bullishness in the stock markets. Brown and Cliff (2004) argue that the VIX index and the II sentiment index are useful proxies for institutional investors’ sentiment. The former is a close approximation because the derivatives market is dominated by institutional investors, and the latter is appropriate because many of the authors of the II newsletters are current or retired professionals. Moreover, the VIX can be considered a forward-looking measure attempting to gauge investors’ fear about the future economy (Simon and Wiggins 2001; Baker and Wurgler 2007; Kurov 2010). Although both the II sentiment index and VIX are commonly used, the latter is preferable. One advantage of the VIX over the II sentiment measure is that the VIX is calculated mechanically using the Black and Scholes option-pricing formula, resulting in the implied volatility, whereas the former has a greater risk of being affected by sampling andmeasurement errors. Lee et al. (2002) use the II survey and finds that the sentimentindex has positive effects on the returns and volatility of the DJIA, S&P500, and NASDAQ. However, they suggest that these results are explained by “noise” traders. More recently, literature shows that investors’ sentiment includes rational and irrational components (Verma and Soydemir 2006). Lee et al. (2002) argue that because the II sentiment index is based on investors’ recommendations, individual investors can be considered as “noise” or irrational traders. Fisher and Statman (2000) point out that the sentiment of Wall Street investors is different from that of individual investors and newsletter writers. Additionally, they conclude that the II survey, which Fisher and Statman use to proxy for newsletter writers’ sentiment, has no statistically significant relationship with future S&P500 returns. Nevertheless, it has the expected negative sign, since the II survey is designed to follow a contrarian strategy, i.e., taking the opposite decision following analysts’ recommendations.

J Econ Finan

Recently, several researchers proxy investor sentiment by employing measures of the implied volatility obtained from the Chicago Board Options Exchange (CBOE) using the Black and Scholes option pricing formula. Corrado and Miller (2005), Wang et al. (2006), and Banerjee et al. (2007) show favorable results after testing the predictive accuracy of the VIX on realized returns. Latane and Rendleman (1976) calculate the weighted implied standard deviation (WISD) using the Black and Scholes formula and find that the WISD is closer to the realized standard deviation than historical estimates. Chiras and Manaster (1978) apply the WISD to determine a trading strategy based on implied volatilities, and they claim that this strategy can earn abnormal returns. Black (2006) mentions that institutional investors, such as hedge fund managers, care about the VIX to optimize portfolios, and hence the VIX can be used to proxy for institutional investor sentiment since option trading is mostly dominated by informed traders. In addition, Bailey et al. (2000) state that institutional investors dominate the ADR market, whereas individual investors dominate the market for closed-end funds. In this study, we use the VIX to proxy for institutional investor sentiment rather than alternative measures. In addition, we select the volatility index because the VIX is calculated on a daily basis, which provides us with more powerful time series.

3 Data The sample includes all level-2 and level-3 ADRs issued by firms in Argentina, Brazil, Chile, and Mexico from January 1995 to May 2009. If an ADR program is cancelled during that period, it is excluded from the sample.5 We divide the data into two subsamples due to a potential structural break in 2002. There are several reasons for this break. First, the Sarbanes-Oxley Act (SOX) was signed into law in July 2002.6 Second, this period coincides with the burst of the dot-com bubble and captures the recovery towards 2008 highs (see Fig. 2). Finally, by breaking the sample on December 31st 2002, we are able to observe one bull and one bear market in each subsample.7 In analyzing data for Latin America, we always have to be cognizant of the impact of various currency crisis episodes. For example, the significant “tequila” effect of the Mexican peso devaluation in 1994 has been well documented (see for example, Kaminsky and Reinhart 1998). This and other currency crisis created shockwaves through the economy, affecting much more than the stock markets, thereby increasing investment uncertainty. The ADR market is therefore often perceived as a way to minimize these fears when investing in Latin American economies. 5

Following the recommendation of an anonymous referee, we note that the sampling procedure may overstate the effect of surviving ADRs. We acknowledge that the results must be interpreted with care. However, we believe the main conclusion is unlikely to change given the robustness of our findings as shown in Tables 3, 4 and 5. 6 SOX was signed into law in July 2002 following a number of high-profile scandals and imposes more stringent disclosure requirements in order to prevent accounting misconduct. Chira (2011) finds that ADR returns have a structural break pre- and post-SOX. 7 This trend can be observed on Fig. 2. Figure 2 employs the CRSP value-weighted index representing the total market. The cutoff point is the month where the stock market reached bottom in 2002.

J Econ Finan

3000 2000 1000

CRSP value weighted

4000

CRSP value-weighted index from January 1995 to May 2009

01 Jan 95

01 Jan 00

01 Jan 05

01 Jun 09

Fig. 2 CRSP value-weighted index from January 1995 to May 2009

We use the VIX as a proxy for institutional investor sentiment.8 Latin American corporations started cross-listing their equity in the early 1990s; however, crosslisting accelerated in 1995.9 In addition, Corrado and Miller (2005) mention that the forecasting ability of the CBOE volatility index, VIX, has improved since 1995. We start our analysis from January 1995 because at least five ADR programs from these Latin American countries were available in the DataStream International database. Latin American ADRs are appropriate because of simultaneous trading activity with the U.S. stocks, and therefore these stock markets should simultaneously react to various stimuli unlike stock exchanges in different regions. The underlying stocks and ADR prices, ADR ratios, exchange rates, and returns on the domestic indices are collected from the DataStream International database. ADR bid-price, ask-price, close-price, and returns on the U.S. stock market (proxied by the Center for Research in Security Prices (CRSP) equal-weighted index) are obtained from the CRSP database. To isolate the consequences of outliers, we winsorize the liquidity and transaction costs variables at the 2 % and 98 % percentile. Descriptive statistics are presented in Table 1. Transaction costs are estimated by using the bid-ask spread and the inverse of the share price, as suggested by Grossman et al. (2007). All countries have positive means in ADR premium, except Mexico, yet the value is close to zero, implying low mispricing. Mexico has the closest value to ADR parity 8

Implied volatility was initially proposed as a measure of expected volatility by Latane and Rendleman (1976) and Chiras and Manaster (1978). 9 Several authors agreed that the mid 1990’s is a period where emerging countries started to become more integrated to world capital markets (Henry 2000; Bekaert and Harvey 2000; Bekaert et al. 2005; Esqueda et al. 2012).

J Econ Finan Table 1 Descriptive statistics Country Argentina

Brazil

Chile

Mexico

U.S.

Variable

N

Mean

Median

Min

Max

S.D 2.28 %

Merval

45120

0.01 %

0.00 %

−15.91 %

14.88 %

Transaction

25087

1.75 %

1.13 %

0.01 %

9.99 %

1.81 %

Premium

30607

0.45 %

0.09 %

−9.89 %

11.00 %

3.63 %

Liquidity

25087

0.57 %

0.49 %

0.14 %

1.83 %

0.35 %

Bovespa

82720

0.04 %

0.03 %

−18.80 %

25.04 %

2.36 %

Transaction

39801

1.00 %

0.44 %

0.01 %

5.88 %

1.27 %

Premium

56207

0.54 %

0.14 %

−35.02 %

35.53 %

5.42 %

Liquidity

39801

1.08 %

0.88 %

0.29 %

4.19 %

0.74 %

IGPA

41360

0.02 %

0.00 %

−5.14 %

8.66 %

0.79 %

Transaction

32088

1.28 %

0.78 %

0.01 %

6.25 %

1.37 %

Premium

33937

1.99 %

0.54 %

−6.36 %

17.47 %

4.20 % 0.81 %

Liquidity

32088

0.96 %

0.73 %

0.16 %

4.32 %

IPC

75200

0.05 %

0.03 %

−15.39 %

11.44 %

1.66 %

Transaction

41857

1.66 %

1.00 %

0.01 %

9.93 %

1.84 %

Premium

53119

−0.16 %

−0.12 %

−14.26 %

27.22 %

5.77 %

Liquidity

41857

0.69 %

0.56 %

0.12 %

2.33 %

0.49 %

CRSP

80388

0.07 %

0.14 %

−8.03 %

10.74 %

0.97 %

VIX

82698

21.33

20.03

9.89

80.86

9.03

S-VIX

82720

21.30

20.15

10.17

72.62

8.76

Benchmark indices: Merval, Bovespa, IGPA, IPC, and CRSP represent the percentage daily returns on the stock market for the corresponding countries. Transactions are proxied by the bid-ask spread and Premium represents the ADR premium. Liquidity is the daily ADR turnover

(−0.16 %). The country with the lowest average transaction costs is Brazil as indicated by a 1 % spread. All values appear to be within reasonable dimensions. Our benchmark indices are the Merval for Argentina, Bovespa for Brazil, Indice General de Precios de Acciones (IGPA) for Chile, and Indice de Precios y Cotizaciones (IPC) for Mexico, respectively. The values of benchmark indices represent the daily returns. All variables are in decimals, except for the Smoothed VIX (S-VIX).10 As expected, the S-VIX shows less volatility than the VIX values. We present the correlation matrix in Table 2. Market capitalization is always negatively correlated with ADR premium, indicating that larger firms are less likely to trade at a premium. This could be due to lower average transaction costs for larger firms; hence, arbitrageurs have more limitations to correct any perceived mispricing. Transaction costs have a positive correlation with ADR premiums in Chile and Mexico; however, this relation appears negative and close to zero in Argentina and

10

The S-VIX is expressed in percentage and follows a smoothing procedure with a 1-month trailing period, where the most recent value has half of the weight.

J Econ Finan

Table 2 Correlation matrix by

S-VIX

country

Transaction

ADR premium

Argentina Transaction Premium MarketCap

0.147*** −0.173*** −.0134**

−0.012* −0.266***

−0.10***

Brazil

The sample is for the period January 1995 to May 2009. ***, **, and * indicate p-values with statistical significance at the 1 %, 5 % and 10 % levels respectively. We calculate the descriptive statistics using daily data for the full sample period. The table includes variables: S-VIX equals the smoothed VIX, Transaction are the transaction costs, Premium is the ADR premium, and MarketCap is the market capitalization

Transaction

0.132***

Premium

−0.095***

−0.087***

MarketCap

−0.018***

−0.263***

−0.023***

Chile Transaction

0.138***

Premium

0.023***

0.25***

MarketCap

0.02***

−0.212***

−0.081

Mexico Transaction

0.282***

Premium

−0.013**

0.230***

MarketCap

−0.092***

−0.245***

−0.03***

Brazil. However, these coefficients must be interpreted carefully given the limitations of bivariate correlations. 4 Econometric methodology We test for stationarity of the series for VIX, ADR premium, and transaction costs by using the Augmented Dickey-Fuller (ADF) test and the Dickey-Fuller generalized least squares (DF-GLS) procedures. We find unit roots in the transactions costs, when they are measured as the inverse of the share price, for all country portfolios. Therefore, we only proxy for transaction costs using the bid-ask spread normalized by the closing price. The ADR premium of stock k at time t is calculated as follows: ADRpremiumkt ¼ ½ADRpricekt −½ADRratiok  Stockpricekt   Exchangeratet ð1Þ ADR premium is standardized and expressed as a percentage of the ADR price. We test the statistical inference between ADR premiums and investor sentiment by using the GARCH-in-mean model (GARCH-M). This methodology is appropriate because there might be a conditional variance of the ADR premium in our sample period due to country events such as currency depreciations. We employ the GARCH-M model, which was put forward by Engle et al. (1987). In this GARCH model, the conditional mean of asset returns is a function of the conditional volatility. We assume that the conditional variance of the realized ADR premium follows the GARCH (1,1) model proposed by Bollerslev (1986). We test the following GARCH-M model where Yt is the ADR premium.

J Econ Finan

Y t ¼ β0 þ β1 σ2tjt−1 þ εt ; σ2 ðεt Þ ¼ Varðεt jΩt Þ ¼ β0 þ

n X

β j ε2t− j þ β 1 σ2tjt−1

ð2Þ ð3Þ

j¼1

Figure 3 shows the time series behavior of the VIX. We observe that this series has frequent fluctuations and can be smoothed. We expect that a conditional variance is present, hence, we employ the S-VIX variable described above.11 4.1 Relation between ADR premium and the U.S. investors’ fear The GARCH-M model allows for asymmetries in the reaction to the increases/decreases in investors’ fear, which fits the nature of our data. The conditional variance is allowed to have an effect on the conditional mean of the dependent variable. The GARCH-M (p, q) models are shown below. The dependent variable, Premium, equals the ADR premium in kth equal-weighted country portfolios of Latin American ADRs. Premiumkt ¼ α þ

n X

εt− j þ

j¼1

n X

VIXt−z þ USindext þ Transcostkt þ σ2 ðεkt Þ þ εkt

z¼1

ð4Þ

εkt ¼ Zt σt Premiumkt ¼ α þ

n X

εt− j þ

j¼1

n X

VIXt−z þ DomestIndext þ USindext

z¼1

ð5Þ

þ Transcosttk þ σ ðεkt Þ þ εkt εkt ¼ Zt σt 2

Premiumkt ¼ α þ

n X j¼1

εt− j þ

n X

VIXt−z þ Liquidtk þ DomestIndext

z¼1

ð6Þ

þ USindext þ Transcosttk þ σ ðεkt Þ þ εkt 2

εkt ¼ Zt σt We expect a negative coefficient for the VIX to support the hypothesis of higher ADR premiums when investors are more optimistic about the stock market. A positive response of the mean term (ARCH-M) indicates that during more volatile periods, the expected value of the dependent variable (ADR premium) is higher.

5 Empirical results Investor sentiment has a strong negative effect on ADR premiums. Results from Eq. (4) indicate a negative relationship between investor sentiment and ADR premiums. Specifically, high levels of VIX (fear) indicate low premiums or discounts. Even after controlling for the U.S. stock-market returns, the lagged S-VIX coefficient is negative 11 When the two models are compared, the latter approach has a higher predictive power than the simple VIX index.

J Econ Finan

0

20

40

60

80

Time-Series line of the Volatility Index from January 1995 to May 2009.

01 Jan 95

01 Jan 00

01 Jan 05 VIX

01 Jun 09

Smoothed_VIX

Fig. 3 Time-series line of the volatility index from January 1995 to May 2009

and significant at the 1 % level in 11 out of the 12 GARCH models presented in Table 3. The exception is in the first subsample for Argentinean ADRs. Perhaps this could be partially attributed to the pegged currency system that Argentina maintained during the whole subsample period for 1995–2002. The U.S. stock market returns positively affect the ADR premium, with higher returns in the U.S. stock market increasing the ADR premium. This is probably due to a higher demand for the foreign asset in days when investors perceive more volatility. Our results are consistent after controlling for the returns on the domestic stock market, using the model from Eq. (5). Table 4 shows similar results to the previous findings in Table 3. The lagged value of S-VIX is negative and significant in all models except for the 1995–2002 subsample of Argentinean ADRs. Additionally, the U.S. stock markets have a positive impact on ADR premiums. Transaction costs are significant and positive in all cases, indicating that deviations from the LOP are at least partially explained by the limits of arbitrage in non-frictionless markets. It seems reasonable to conclude that the previous 1-day value of the VIX incorporate useful information in predicting Latin American ADR premiums. AnegativeARCH-McoefficientindicatesthatmorevolatilemarketstriggerlowerADR premiums, i.e. reduce the premium or widen the discount. The direction of the coefficient differsacrosscountries.ArgentineanADRstendtolowerthepremiuminmorevolatiledays inthemarket,asindicatedbythenegativeARCH-Mcoefficient.On the contrary, Chilean ADRs increase the premium during more volatile periods. The different reactions might be due to investors perceiving Chilean ADRs as safer, since Chile has crosslisted mining and utility companies which are generally perceived as less risky firms. We find that Mexico and Brazil do not have significant ARCH-M coefficients; however, when we split the sample, the terms become significant but in opposite directions. Surprisingly, Mexican and Brazilian ADRs respond differently to volatility depending on market conditions. For example, in the more recent period (2003–2009), investors tend to pay in excess of parity for Mexican and

J Econ Finan n

n

Table 3 Results from GARCH-M, Eq. (4) base model Premiumkt ¼ α þ ∑ εt− j þ ∑ VIXt−z þ j¼1 z¼1 USindext þ Transactkt þ σ2 ðεkt Þ þ εkt Argentina

Brazil

Chile

Mexico

S-VIX

−0.170***

−0.062***

−0.025***

−0.021***

CRSP

0.144***

0.153***

0.179***

0.048***

Panel A: full sample

Transaction

0.425***

−1.029*** −0.695

0.125***

0.511***

ARCH-M

−8.020***

ARCH(1)

0.658***

0.561***

0.254***

0.509***

GARCH(1)

0.381***

0.401***

0.750***

0.446***

29.213***

0.148

N=3524 Panel B: subsample: 1995–2002 S-VIX

−0.007

−0.115***

−0.095***

−0.023***

CRSP

−0.026

−0.042

0.119***

−0.052***

0.167***

−0.817***

0.008

−0.228***

−0.528

Transaction ARCH-M

3.701***

11.450***

−8.891***

ARCH(1)

0.753***

0.673***

0.516***

0.448***

GARCH(1)

0.269***

0.297***

0.440***

0.443***

−0.023***

N=2013 Panel C: subsample: 2003–2009 S-VIX

−0.169***

−0.063***

−0.025***

CRSP

0.226***

0.230***

0.138***

Transaction

0.096*

0.077

0.233***

0.012 0.337***

ARCH-M

−6.791***

8.696***

27.164***

ARCH(1)

0.521***

0.388***

0.198***

0.657***

GARCH(1)

0.395***

0.574***

0.768***

0.297***

3.235*

N=1511 Dependent Variable is ADR premium. Panel A includes the complete sample period. Panel B includes from January 1995 to December 2002. Panel C is the subsample for the period of January 2003 to May 2009. ***, **, and * indicate p-values with statistical significance at the 1 %, 5 % and 10 % levels respectively

Brazilian ADRs, but only when the U.S. stock markets were perceived to be more volatile. The U.S. stock market is positive and significant in most cases, validating the hypothesis that ADR premiums are affected by the host stock-market performance. ARCH (p) and GARCH (q) coefficients are significant at the 1 % level in all cases, indicating that this methodology is appropriate; this is probably due to substantial volatility increases during the financial crises in Argentina (2001–2002), Brazil (1999, 2002, and 2008), Chile (2008), and Mexico (1994, 1997, and in 2008).12 During the sample period, there was an important expansion in Latin American economies that can be reflected in liquidity growth. Grossman et al. (2007) control for liquidity measures and Baker and Wurgler (2007) describe its relevance in contemporary research. 12

See Esqueda and Jackson (2012) for a description of the financial crises in these four Latin American economies.

J Econ Finan n

Table 4 Results from GARCH-M, Eq. (5) domestic index control variable Premiumkt ¼ α þ ∑ n

j¼1

εt− j − þ ∑ VIXt−z þ DomestIndext þ USindext þ Transacttk þ σ2 ðεkt Þ þ εkt z¼1

Argentina

Brazil

Chile

Mexico

Panel A: full period Domestic index

−0.084***

−0.151***

−0.109***

−0.067***

S-VIX

−0.168***

−0.074***

−0.025***

−0.021***

CRSP

0.336***

0.320***

−0.109***

0.113***

Transaction

0.403***

−1.020***

0.124***

0.507***

ARCH-M

−8.089***

ARCH(1)

0.664***

−0.642 0.575***

0.256***

0.541***

GARCH(1)

0.382***

0.381***

0.748***

0.405***

29.160***

−0.085

N=3524 Panel B: subsample: 1995–2002 Domestic index

−0.053***

−0.166***

0.005***

−0.025**

S-VIX

−0.007

−0.071***

−0.095***

−0.022***

CRSP

−0.038

Transaction

0.137***

0.118***

−0.024

0.177***

−0.131***

0.008***

−0.230***

ARCH-M

3.695***

−2.121***

11.470***

−8.771***

ARCH(1)

0.756***

0.785***

0.515***

0.449***

GARCH(1)

0.263***

0.237***

0.441***

0.443***

N=2013 Panel C: subsample: 2003–2009 Domestic index

−0.115***

−0.117***

−0.138***

−0.047***

S-VIX

−0.170***

−0.071***

−0.024***

−0.023***

CRSP

0.352***

0.382***

0.174***

0.062***

Transaction

0.080**

0.151***

0.217***

0.334***

ARCH-M

−7.514***

0.971***

25.080***

3.228***

ARCH(1)

0.485***

0.321***

0.221***

0.665***

GARCH(1)

0.459***

0.651***

0.747***

0.284***

N=1511 Dependent Variable is ADR premium. Panel A includes the complete sample period. Panel B includes from January 1995 to December 2002. Panel C is the subsample for the period of January 2003 to May 2009. ***, **, and * indicate p-values with statistical significance at the 1 %, 5 % and 10 % levels respectively. Domestic indices are: Merval, Bovespa, IGPA, IPC, from the Argentinean, Brazilian, Chilean, and Mexican stock markets respectively

Hence, in Eq. 6 we include a variable (turnover) that proxies for liquidity.13 The results are presented in Table 5. Turnover is included to control for potential differences in ADR liquidity and their likely effect on ADR premiums. We find that S-VIX remain significant in all model specifications, albeit control variables such as domestic index and transaction costs slightly reduce their contribution to the models for the 1995–2002 subsample. The 13

Turnover has been commonly estimated as the volume of shares traded divided by the number of shares outstanding. We estimate daily turnover and is winsorized at the 2 % and 98 % levels.

J Econ Finan Table 5 Results from GARCH-M, Eq. (6) domestic index & liquidity control variables Premiumkt ¼ n

n

j¼1

z¼1

α þ ∑ εt− j þ ∑ VIXt−z þ Liquidtk þ DomestIndext þ USindext þ Transacttk þ σ2 ðεkt Þ þ εkt Argentina

Brazil

Chile

Mexico

Panel A: full period Domestic index

−0.087***

−0.153***

−0.089***

−0.065***

S-VIX

−0.531***

−0.072***

−0.029***

−0.020***

U.S. index/CRSP

0.535***

0.336***

0.225***

0.110***

Transaction

1.323***

−1.013***

0.617***

0.472***

0.125***

−0.017

−0.171***

Liquidity

0.781***

ARCH-M

−1.890***

ARCH(1)

0.707***

0.568***

0.618***

0.544***

GARCH(1)

0.299***

0.394***

0.637***

0.392***

−0.591

1.473***

−0.194

N=3524 Panel B: subsample: 1995–2002 Domestic index

−0.028*

−0.169***

−0.001

−0.023

S-VIX

−1.573***

−0.066***

−0.094***

−0.016***

U.S. index/CRSP

0.402***

0.135***

Transaction costs

0.225***

−0.138***

0.119***

−0.024

0.008

−0.233***

Liquidity

0.091

0.173***

0.073***

0.285***

ARCH-M

0.327**

−2.186***

11.349***

−9.837***

ARCH(1)

0.911***

0.766***

0.518***

0.429***

GARCH(1)

0.117***

0.254***

0.439***

0.472***

N=2013 Panel C: subsample: 2003–2009 Domestic index

−0.116***

−0.112***

−0.077***

−0.046***

S-VIX

−0.168***

−0.069***

−0.039***

−0.023***

U.S. index/CRSP

0.351***

0.370***

0.115***

0.061**

Transaction costs

0.084

0.160

Liquidity

0.349***

−0.060

−0.124* 0.125*** −0.233

0.328*** −0.015

ARCH-M

−7.572***

9.943***

ARCH(1)

0.491***

0.313***

1.190***

0.666***

GARCH(1)

0.461***

0.666***

0.267***

0.283***

3.194**

N=1511 Dependent Variable is ADR premium. Panel A includes the complete sample period. Panel B includes from January 1995 to December 2002. Panel C is the subsample for the period of January 2003 to May 2009. ***, **, and * indicate p-values with statistical significance at the 1 %, 5 % and 10 % levels respectively. Domestic indices are: Merval, Bovespa, IGPA, IPC, from the Argentinean, Brazilian, Chilean, and Mexican stock markets respectively

main conclusions remain robust after incorporating the liquidity effects, as the coefficient of the S-VIX variable remains negative and significant. We therefore conclude that implied volatility generated from option prices can be employed to trade ADR the following day, even after controlling for market frictions.

J Econ Finan

6 Conclusions We confirm the intuitive presumption that when the U.S. stock market is doing well, investors tend to get overly optimistic, as suggested by the positive and significant coefficients of the U.S. index variable. Moreover, transaction costs are strongly significant. The results support the notion that deviations from the LOP in the ADR market are partially due to the limits of arbitrage in non-frictionless markets. The VIX has a significant negative impact on ADR premiums of Latin American countries. When institutional investors believe that stock market volatility is about to rise, ADR premiums tend to decrease. We find that deviations from the LOP in ADRs can be predicted by the lagged values of the volatility index. We present evidence that during our most recent subsample, when volatility in the ADR market increases, deviations from the LOP increase as well. This is true for ADRs from Chile, Mexico, and Brazil, whereas the opposite view holds for Argentinean ADRs. This paper has important implications for ADR investors, since they can improve diversification and hedging strategies by incorporating known values of the volatility index. Finally, we hypothesize that investors might not react symmetrically to innovations about the economy because investors have different expectations in bull and bear markets. Nelson (1991) states that news arriving to the market has an asymmetric impact on stock volatility. Lee et al. (2002) find that bad news has an asymmetric effect on conditional volatility and as sentiment becomes more bearish, there is a potential reduction in excess returns. Investors’ outlook about the economy can have a different outcome on the equilibrium prices of ADRs. However, whether the implied volatility causes an asymmetric response on ADR mispricing remains open for future research.

References Alhaj-Yaseen YS (2011) Cross-listing in the home market after going public in the U.S. J Econ Financ 1–19 Aquino K, Poshakwale S (2006) Price determinants of American Depositary Receipts (ADR): a crosssectional analysis of panel data. Appl Financ Econ 16(16):1225–1237 Bae SC, Kwon TH, Li M (2008) Foreign exchange rate exposure and risk premium in international investments: evidence from American depositary receipts. J Multinatl Financ Manag 18(2):165–179 Bailey W, Chan K, Chung YP (2000) Depositary receipts, country funds, and the peso crash: the intraday evidence. J Financ 55:2693–2717 Baker M, Wurgler J (2006) Investor sentiment and the cross section of stock returns. J Financ 61(4):1645–1680 Baker M, Wurgler J (2007) Investor sentiment in the stock market. J Econ Perspect 21(2):129–151 Banerjee P, Doran J, Peterson D (2007) Implied volatility and future portfolio returns. J Bank Financ 31:3183– 3199 Barberis N, Thaler R (2003) Handbook of the economics of finance. A survey of behavioral finance. Elsevier, vol. 1 chapter 18: 1051–1121 Bekaert G, Harvey C (2000) Foreign speculators and emerging equity markets. J Financ 55(2):565–613 Bekaert G, Harvey C, Lundblad C (2005) Does financial liberalization spur growth? J Financ Econ 77:3–55 Bin FS, Blenman LP, Chen DH (2004) Valuation impact of currency crises: evidence from the ADR market. Int Rev Financ Anal 13(4):411–432 Black K (2006) Improving hedge fund risk exposures by hedging equity market volatility, or how the VIX ate my kurtosis. J Trading 1(2):6–15 Bollerslev T (1986) Generalized autoregressive conditional heteroskedasticity. J Econ 31:307–327 Brown G, Cliff M (2004) Investor sentiment and the near-term stock market. J Empir Financ 11:1–27

J Econ Finan Chira I (2011) The impact of governance characteristics on the stock price of cross-listed companies. J Econ Financ 1–18 Chiras D, Manaster S (1978) The information content of option prices and a test of market efficiency. J Financ Econ 6(2–3):213–223 Corrado C, Miller TW (2005) The forecasting quality of CBOE implied volatility indexes. J Futur Mark 25(4):339–373 DeLong J, Shleifer A, Summers L, Waldmann R (1990) Positive feedback investment strategies and destabilizing rational speculation. J Financ 45(2):379–395 Dennis P, Mayhew S (2002) Risk-neutral skewness: evidence from stock options. J Financ Quant Anal 37(3):471–493 Engle RF, Lilien DM, Robins RP (1987) Estimation of time varying risk premia in the term structure: the ARCH-M model. Econometrica 55:391–407 Esqueda O, Jackson D (2012) Currency depreciation effects on ADR returns: evidence from Latin America. J Econ Financ 36(3):691–711 Esqueda O, Assefa T, Mollick AV (2012) Financial globalization and stock market risk. J Int Financ Mark Inst Money 22(1):87–102 Fama E (1970) Efficient capital markets: a review of theory and empirical work. J Financ 25(2) Fang H, Loo J (2002) Pricing of American depository receipts under market segmentation. Glob Financ J 13:237–252 Fehr E, Tyran J (2005) Individual irrationality and aggregate outcomes. J Econ Perspect 19(4):43–66 Fisher K, Statman M (2000) Investor sentiment and stock returns. Financ Anal J 56(2):16–23 Grossman A, Ozuna T, Simpson MW (2007) ADR mispricing: do costly arbitrage and consumer sentiment explain the price deviation. J Int Financ Mark Inst Money 17(4):361–371 Henry P (2000) Stock market liberalization, economic reform, and emerging market equity prices. J Financ 55(2):529–564 Jackson D, Madura J (2003) Profit warnings and the pricing behavior of ADRs. J Behav Financ 4(3):131– 136 Kaminsky GL, Reinhart CM (1998) Financial crisis in Asia and Latin America: then and now. Am Econ Rev 88(2):444–448 Kim M, Szakmary AC, Mathur I (2000) Price transmission dynamics between ADRs and their underlying foreign securities. J Bank Financ 24:1359–1382 Kurov A (2010) Investor sentiment and the stock market’s reaction to monetary policy. J Bank Financ 34(1):139–149 Latane H, Rendleman R (1976) Standard deviations of stock price ratios implied in option prices. J Financ 31(2) Lee C, Shleifer A, Thaler T (1991) Investor sentiment and the closed-end fund puzzle. J Financ 46:75–109 Lee W, Jiang C, Indro D (2002) Stock market volatility, excess returns, and the role of investor sentiment. J Bank Financ 26:2277–2299 Liang Y, Mougoué M (1996) The pricing of foreign exchange risk: evidence from ADRs. Int Rev Econ Financ 5(4):377–385 Nelson DB (1991) Conditional heteroskedasticity in asset returns: a new approach. Econometrica 59(2):347–370 Pontiff J (1996) Costly arbitrage: evidence from closed-end funds. Q J Econ 111(4):1135–1151 Shleifer A, Vishny R (1997) The limits of arbitrage. J Financ 52(1):35–55 Simon DP, Wiggins RA (2001) S&P futures returns and contrary sentiment indicators. J Futur Mark 21(5):447–462 Suarez D (2005) Arbitrage opportunities in the depositary receipts market: myth or reality? J Int Financ Mark Inst Money 15:469–480 Verma R, Soydemir G (2006) The impact of U.S. individual and institutional investor sentiment on foreign stock markets. J Behav Financ 7(3):128–144 Verma R, Soydemir G (2009) The impact of individual and institutional investor sentiment on the market price of risk. Q Rev Econ Financ 49(3):1129–1145 Verma P, Verma R (2008) Are survey forecasts of individual and institutional investor sentiments rational? Int Rev Financ Anal 17(5):1139–1155 Wang Y, Keswani A, Taylor S (2006) The relationships between sentiment, returns, and volatility. Int J Forecasting 22:109–123